FeatureExtraction.py

import cv2
import numpy as np
from tensorflow.keras.applications.resnet import ResNet152
from tensorflow.keras.layers import AveragePooling2D, Flatten
from tensorflow.keras.models import Model

class FeatureExtractor:
    def __init__(self, img_shape, channels, seq_length):
        self.seq_length = seq_length
        self.height = img_shape[0]
        self.width = img_shape[1]
        self.channels = channels
        self.base_model = ResNet152(include_top=False, input_shape=(224, 224, 3), weights='imagenet')
        for layer in self.base_model.layers:
            layer.trainable = False
        self.op = self.base_model.output
        self.x_model = AveragePooling2D((7, 7), name='avg_pool')(self.op)
        self.x_model = Flatten()(self.x_model)
        self.model = Model(self.base_model.input, self.x_model)

    def extract_feature(self, frames_buffer):
        x_op = np.zeros((2048, self.seq_length))
        for i in range(len(frames_buffer)):
            x_t = frames_buffer[i]
            x_t = cv2.resize(x_t, (224, 224))
            x_t = np.expand_dims(x_t, axis=0)
            x = self.model.predict(x_t)
            x_op[:, i] = x.flatten()
        return x_op